Sort the four statements into proven indications versus the one that the evidence rejects. Three describe situations where antibiotics genuinely help, and one describes blanket prophylaxis that trials have shown to be useless.
The odd one out is giving antibiotics simply because a woman has threatened preterm labour with her membranes still intact. The ORACLE II study tested exactly this and found no improvement in outcomes, plus a hint of long-term harm to the children. Routine prophylaxis with unruptured membranes is therefore not recommended, which makes that statement incorrect.
The other three are accepted practice. Bacterial vaginosis, even without symptoms, is linked to preterm birth, so metronidazole is appropriate. Asymptomatic bacteriuria in pregnancy is always treated because it can progress to pyelonephritis and trigger preterm labour. And once the membranes rupture early (PPROM), antibiotics such as erythromycin prolong the pregnancy and cut infection rates.
So the statement that is not correct is the one advising prophylaxis with unruptured membranes.
\[\boxed{\text{Antibiotic prophylaxis even with unruptured membranes}}\]